Provided is an observation condition determination support device which can improve the defect classification accuracy. The observation condition determination support device includes: a means (26) for acquiring a plurality of defects images which have captured the same defect under a plurality of observation conditions set in advance in an observation device (5) in accordance with check data relating to defects of a semiconductor device detected by an inspection device (4); a means (12) for classifying the plurality of the same defects according to the respective defect images and determining a first category to which the same defects belong for each of the observation conditions as a result of the classification; and a means (13) for determining an observation condition to be used when fabricating the semiconductor device among the plurality of the observation conditions according to the ratio at which the first category is matched with a second category determined by a user of the observation device who has classified the same defects.
Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. An observation condition determination support device, comprising: means for acquiring a plurality of defect images of a same defect photographed under a plurality of observation conditions set in advance in an observation device, based on an inspection data of the same defect of a semiconductor device detected by an inspection device; means for determining a first category to which the same defect belongs for each of the plurality of the observation conditions as a result of classification, by classifying a plurality of the same defects based on each of the defect images; and means for determining the observation condition to be used when the semiconductor device is fabricated from the plurality of the observation conditions, based on a rate that the first category is matched with a second category which is determined by a user of the observation device by classifying the same defects.
The observation condition determination support device helps determine the best settings on an observation device for detecting defects in semiconductor manufacturing. The device first acquires multiple images of the same defect, where each image is taken under different predefined observation settings. Using these images, the device automatically classifies each defect under each observation setting. Then, it compares these automatic classifications to classifications made by a human operator. Based on how well the automated classifications match the human classifications for each observation setting, the device recommends the best observation setting to use during manufacturing.
2. The observation condition determination support device according to claim 1 , further comprising: means for displaying a list of the second category and the first category corresponding to each of the plurality of the observation conditions, for each of the same defects.
In addition to the features described in claim 1, this observation condition determination support device also displays a list showing both the automated classification ("first category") and the human expert's classification ("second category") for each of the different observation conditions and each defect. This allows the user to easily compare the machine's performance against their own for each condition, enabling a more informed decision about the optimal observation settings.
3. The observation condition determination support device according to claim 1 , further comprising: means for displaying a list of a plurality of the defect images of the same defect photographed under the plurality of the observation conditions, for each of the same defects.
In addition to the features described in claim 1, this observation condition determination support device also displays a list of the images of the same defect, where each image was captured using a different observation condition. This visual comparison helps the user understand how different observation settings affect the appearance of defects, contributing to more accurate defect classification and determination of ideal observation parameters.
4. The observation condition determination support device according to claim 3 , wherein the means for displaying a list of a plurality of the defect images displays a list of a plurality of the defect images and requires input of the second category determined by the user.
Building upon the features from claim 3 where multiple defect images under different observation conditions are displayed, this version of the observation condition determination support device requires the user to input their classification (the "second category") directly while viewing the displayed images. This integrated workflow streamlines the process of collecting human expert classifications, facilitating comparison with the automated classification results.
5. The observation condition determination support device according to claim 1 , wherein the means for acquiring a plurality of defect images skips acquisition of the defect images of a defect on a coordinate identical to the coordinate of the defect detected by an inspection process before a preceding inspection process.
The observation condition determination support device as described in claim 1 optimizes image acquisition by skipping image capture for defects that have already been detected and imaged in a prior inspection step at the same location. This reduces redundant data collection and processing, making the overall defect analysis process faster and more efficient.
6. The observation condition determination support device according to claim 1 , wherein the means for acquiring a plurality of defect images determines a rate that the semiconductor device becomes defective by the defect by using a layout data of the semiconductor device, and based on the rate, extracts the defect as a defect for acquiring the defect images.
The observation condition determination support device as described in claim 1 uses the layout data of the semiconductor device to determine the probability that a particular defect will cause the device to fail. Based on this probability, the device prioritizes the acquisition of defect images for those defects most likely to be critical, ensuring that the most important defects are analyzed first.
7. The observation condition determination support device according to claim 1 , wherein the means for acquiring a plurality of defect images classifies the defect based on a distribution of the defect within a wafer which is formed by coordinates of a plurality of the defects, and extracts the defect from the classified defects as a defect for acquiring the defect images.
The observation condition determination support device as described in claim 1 analyzes the spatial distribution of defects across the semiconductor wafer and groups them into classes based on their location patterns. It then selects representative defects from each class for image acquisition and analysis. This ensures that the analysis covers a diverse range of defect types and locations, improving the overall accuracy of the observation condition determination process.
8. The observation condition determination support device according to claim 1 , further comprising: means for extracting a defect whose first category is not matched with the second category; and means for requiring input for changing the first category or the second category, while displaying together the defect images of defects having a category identical to the first category of the extracted defect and the defect images of defects having a category identical to the second category of the extracted defect.
Expanding on the functionality in claim 1, this version of the observation condition determination support device identifies defects where the automated classification (first category) does not match the user's classification (second category). The system then displays the images of these mismatched defects alongside images of other defects that *do* match each category. This allows the user to visually compare and re-evaluate the classifications, and correct either the automated or manual classification as needed, refining the system's overall accuracy.
9. An observation condition determination support method, in which a computer executes steps of: acquiring a plurality of defect images of a same defect photographed under a plurality of observation conditions set in advance in an observation device, based on an inspection data of the same defect of a semiconductor device detected by an inspection device; determining a first category to which the same defect belongs for each of the plurality of the observation conditions as a result of classifying, by classifying a plurality of the same defects based on each of the defect images; and determining the observation condition to be used when the semiconductor device is fabricated from the plurality of the observation conditions, based on a rate that the first category is matched with a second category which is determined by the user of the observation device by classifying the same defects.
The observation condition determination support method is a computer-implemented process for determining the optimal observation settings for detecting semiconductor defects. First, the process acquires multiple images of the same defect, captured under different predefined observation conditions. Second, it classifies each defect automatically under each observation setting. Third, it compares these automated classifications to classifications provided by a human operator. Finally, based on how well the automated classifications match the human classifications, the method determines the best observation setting to use.
10. The observation condition determination support method according to claim 9 , further comprising a step of: displaying a list of the second category and the first category corresponding to each of the plurality of the observation conditions, for each of the same defects.
In addition to the method described in claim 9, this method also displays a list showing both the automated classification ("first category") and the human expert's classification ("second category") for each of the different observation conditions and each defect. This allows the user to easily compare the machine's performance against their own for each condition, enabling a more informed decision about the optimal observation settings.
11. The observation condition determination support method according to claim 9 , further comprising a step of: displaying a list of the plurality of the defect images of the same defect photographed under the plurality of the observation conditions, for each of the same defects.
In addition to the method described in claim 9, this method also displays a list of the images of the same defect, where each image was captured using a different observation condition. This visual comparison helps the user understand how different observation settings affect the appearance of defects, contributing to more accurate defect classification and determination of ideal observation parameters.
12. The observation condition determination support method according to claim 11 , wherein in the displaying a list of the plurality of the defect images, displaying a list of a plurality of the defect images; and requiring input of the second category determined by the user.
Building upon the method from claim 11 where multiple defect images under different observation conditions are displayed, this version of the observation condition determination support method requires the user to input their classification (the "second category") directly while viewing the displayed images. This integrated workflow streamlines the process of collecting human expert classifications, facilitating comparison with the automated classification results.
13. The observation condition determination support method according to claim 9 , wherein in the acquiring a plurality of defect images, skipping acquisition of the defect images of a defect on a coordinate identical to the coordinate of the defect detected by an inspection process before a preceding inspection process.
The observation condition determination support method as described in claim 9 optimizes image acquisition by skipping image capture for defects that have already been detected and imaged in a prior inspection step at the same location. This reduces redundant data collection and processing, making the overall defect analysis process faster and more efficient.
14. The observation condition determination support method according to claim 9 , wherein in the acquiring a plurality of defect images, determining a rate that the semiconductor device becomes defective by the defect by using a layout data of the semiconductor device; and based on the rate, extracting the defect as a defect for acquiring the defect images.
The observation condition determination support method as described in claim 9 uses the layout data of the semiconductor device to determine the probability that a particular defect will cause the device to fail. Based on this probability, the method prioritizes the acquisition of defect images for those defects most likely to be critical, ensuring that the most important defects are analyzed first.
15. The observation condition determination support method according to claim 9 , wherein in the acquiring a plurality of defect images, classifying the defect based on a distribution of the defect within a wafer which is formed by coordinates of a plurality of defects; and extracting the defect from the classified defects as a defect for acquiring the defect images.
The observation condition determination support method as described in claim 9 analyzes the spatial distribution of defects across the semiconductor wafer and groups them into classes based on their location patterns. It then selects representative defects from each class for image acquisition and analysis. This ensures that the analysis covers a diverse range of defect types and locations, improving the overall accuracy of the observation condition determination process.
16. The observation condition determination support method according to claim 9 , further comprising steps of: extracting a defect whose first category is not matched with the second category; and requiring input for changing the first category or the second category, while displaying together the defect images of defects having a category identical to the first category of the extracted defect and the defect images of the defects having a category identical to the second category of the extracted defect together.
Expanding on the method in claim 9, this version of the observation condition determination support method identifies defects where the automated classification (first category) does not match the user's classification (second category). The system then displays the images of these mismatched defects alongside images of other defects that *do* match each category. This allows the user to visually compare and re-evaluate the classifications, and correct either the automated or manual classification as needed, refining the system's overall accuracy.
Cooperative Patent Classification codes for this invention. Click any code to explore related patents in that topic.
November 19, 2009
June 25, 2013
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.